README.md

Using Twitter to understand psychological responses to #MeToo

This repository is the analysis pipeline for a project looking at how the perceived morality of public figures changes after being accused of sexual misconduct. All cleaned/prepared data ready for analysis, AND raw data, can be found on the OSF page. All scripts run in Python 3.7.6.

Analysis procedures

1. Run cleaning.py to clean and concatenate all of the tweets in the tweets folder on OSF.
2. Run MFD.py to analyze the moral content of tweets.
3. Run afinn_positivity.py and st_positivity.py to calculate how well-liked public figures were, using both a dictionary-based method (AFINN) and a machine learning method (BERT, via simple transformers). Note that simple transformers works best in a conda environment. Set one up using the instructions here.
4. Run the "clean" section analyses.R to load other associated data and combine it into a single dataframe.
5. You can recreate analyses by running the rest of analyses.R, or you may load all models by opening models.RData